Académique Documents
Professionnel Documents
Culture Documents
Senol Emir
Corresponding Author: Asst. Prof., Faculty of Economics, Istanbul University, 34126 Beyazit,
Istanbul, Turkey
Hasan Dincer
Assoc.Prof. of Finance, Istanbul Medipol University, School of Business and Management,
Beykoz, 34810, Istanbul, Turkey
Umit Hacioglu
Assoc.Prof. of Finance, Istanbul Medipol University, School of Business and Management,
Beykoz, 34810, Istanbul, Turkey
Serhat Yuksel
Asst.Prof. of Economics & Finance, Konya Food & Agriculture University, Faculty of Social
Sciences and Humanities, Konya, Turkey
Abstract
In a data set, an outlier refers to a data point that is considerably different from the others. Detecting outliers
provides useful application-specific insights and leads to choosing right prediction models. Outlier detection
(also known as anomaly detection or novelty detection) has been studied in statistics and machine learning
for a long time. It is an essential preprocessing step of data mining process. In this study, outlier detection
step in the data mining process is applied for identifying the top 20 outlier firms. Three outlier detection
algorithms are utilized using fundamental analysis variables of firms listed in Borsa Istanbul for the 2011-
2014 period. The results of each algorithm are presented and compared. Findings show that 15 different
firms are identified by three different outlier detection methods. KCHOL and SAHOL have the greatest
number of appearances with 12 observations among these firms. By investigating the results, it is concluded
that each of three algorithms makes different outlier firm lists due to differences in their approaches for
outlier detection.
45
Key Words: Outlier Detection, Fundamental Analysis, Stock Exchange, k-Nearest Neighbor (k-NN) Global
Outlier Score, Histogram Based Outlier Score (HBOS), Robust Principal Component Analysis (rPCA)
Page
Outlier Score.
Emir, et al. / International Journal of Finance & Banking Studies
Vol 4, No 4, 2015 ISSN: 2147-4486
Introduction
Financial Analysis is the process of evaluating financial reports of the companies (Cooper and Kaplan, 1991). In this
process, financial experts compare some important accounts over the years in order to make comment about financial
performance of them. With the help of financial analysis, it can be possible to define future performance of these
companies. Therefore, financial reports of the companies are so significant that they affect the investment decisions for
these companies.
Ratio analysis is the main part of the financial analysis. In this process, some ratios of the company, created by using
financial accounts, are evaluated (Nissim and Penman, 2001). Hence, it is used to determine extraordinary changes for
some accounts over the years. Another benefit of ratio analysis is that it defines the relationship between some important
accounts. As a result, ratio analysis helps to make comment about financial position of that company.
Because of the advantages emphasized above, ratio analysis is very helpful in order to make investment decision. Many
investors use financial ratios of the company to learn the financial situation. In other words, these ratios support investors
to evaluate financial performance of the company. Therefore, it can be said that ratio analysis helps investors to make
reasonable decision and to minimize the risk in this process (Johnson, 1970).
However, there are some criticisms about the quality of ratio analysis. According to efficient market hypothesis,
necessary information in order to gain profit from the stocks can be reached by everybody (Fama, 1998). Therefore, it
is impossible to get profit by using ratio analysis related to the company. The main reason behind this situation is that
financial reports of the companies may not reflect the real condition. Hence, ratios created from these reports may give
misleading results.
The paper is organized as follows: second part explains fundamental analysis method. The third part gives information
about outlier detection. The fourth part includes empirical analysis. Finally, the analyze results were given at conclusion.
Literature Review
There are many studies in the literature related to fundamental analysis. Most of them analyze the performance of
fundamental analysis. Some of these studies are explained in the table 1.
Shen and Tzeng tried to create a soft computing model for stock selection by using fundamental analysis. This model
included 20 different rules in order to classify the value of the stocks. In addition to them, stocks in Taiwan were used
in this study to achieve this objective. As a result, it was defined that fundamental analysis is very helpful in order to
predict stock price (Shen and Tzeng, 2015). Bistrova and Lace tested the relevance of fundamental analysis on Baltic
equity markets. Within this scope, the data of 45 companies for the period between 2000 and 2008 was used. As a result,
of the analysis, it was determined that fundamental analysis is not helpful in order to get profit from the stock investment
in Baltic market (Bistrova and Lace, 2015). Yan and Zheng made a study about the relationship between fundamental
analysis and stock returns. Within this context, monthly data for the period between 1963 and 2013 was used in this
study. Furthermore, they defined 217 different ratios in this analysis. Finally, they concluded that fundamental signals
are significant predictors of stock returns (Yan and Zheng, 2015). Shukla and Nerlekar tried to determine which tools
of fundamental analysis are successful in determining the stock price. Within this scope, sugar sector in India was
analyzed in this study. Moreover, monthly data between 2008 and 2012 and 7 different ratios related to fundamental
analysis was used. It was concluded that inventory turnover ratio is the key indicator for sugar sector of India (Shukla
and Nerlekar, 2015).
46
Page
to fundamental analysis result, it was determined that ICICI Bank has the greatest point (Megharaja, 2015). Goodman
Page
and others tried to analyze whether fundamental analysis can predict absolute stock price movements or not in option
market. To achieve this purpose, the data between the years 1996 and 2001 was used in this study. As a result, it was
to GDP growth rate, unemployment rate should also be taken into the consideration. In case of high unemployment,
there can be economic recession in the country. Moreover, high inflation rates decrease stability in an economy. That is
Page
to say, when there is low inflation rate in the country, because of stable economic environment, investors become more
Liquidity ratios show how liquid the company is for a given period. In other words, they give information about the
solvency of the company for short-term debt (Saleem and Rehman, 2011). Some ratios related to the liquidity are
emphasized below.
Current Ratio is the ratio of current assets to short-term liabilities. As it can be understood from this definition, this ratio
shows the capacity of the company to pay short-term debts. This ratio is expected to be more than two, but it is also
recommended to evaluate this ratio according to the sector of the company (Edmister, 1972). Liquidity Ratio gives
information about the capacity of the company to pay its short-term debt without using its stock. It is accepted that if
this ratio is greater than 1, it means that this company can pay its short-term debt easily (Howton and Perfect, 1998).
Cash Ratio shows the capacity of the company to pay its debt in an emergency condition. It is accepted for a company
49
to have cash ratio higher than 0.2 in order not to have any problem (Callao, et. al., 2007).
Page
using predictive methods and sampling techniques that are robust to outliers affects the quality of results considerably.
Page
Empirical Analysis
The analysis is constructed using the data set from 2011 to 2014. The end of year values are considered to analyze the
fundamental financial parameters. Total number of observations is 400 for initial raw data set. Then, observations that
have missing values are excluded and final size of data set is reduced to 349.
In the literature, there are several kinds of fundamental analysis parameters to evaluate firm values. However, the
limitation of this study is to consider common parameters that fit to all firms and Table 3 gives information on the
selected variables for creating outlier detection models. Accordingly, nine fundamental variables are employed to
evaluate the outliers.
51
Page
Variable Symbol
Current Ratio V1
Cash Ratio V2
Net Sales V3
Net Profit V4
Total Assets V5
Short Term Liabilities V6
Net Profit / Net Sales V7
Net Profit / Total Assets V8
Short Term Liabilities / Total Assets V9
Descriptive of variables is given in Table 4. Minimum, maximum and mean values of variables are presented. These
values provide useful information for evaluating the results of outlier detection models.
Table 4: Descriptive Statistics for model variables
Table 5 illustrates top 20 outlier results of k-NN Global Outlier Detection algorithm. Nine different firms take place in
the outlier list. The top outlier observation (SAHOL-31.12.2014) has the maximum value of V5 (230.526.555.000) and
V6 (169.096.825.000). It has greater values of V3, V4 and V9 than means of these variables. SAHOL is in first four
place, KCHOL and TUPRS have the next four places consecutively. THYAO follows these firms with two observations.
EKGYO, YAZIC, BIMAS, TCELL and AEFES are the other outliers in the list. K-NN global outlier detection algorithm
commonly identifies as outlier the holdings and companies with large sales volume, short-term liabilities and net profit.
Table 6 figures out the top 20 outlier results of Histogram - Based Outlier Detection algorithm. The top outlier
observation (SAHOL-31.12.2013) in this algorithm is also the first outlier in k-NN. The top eight seats are shared by
SAHOL and KCHOL with four observations each. Histogram Based method does not produce a consecutive outlier
order structure as k-NN as expected. Eight different firms take place in the outlier list. VKGYO, KOZAL, ECZYT are
firms in the list that do not appear in the k-NN results.
Histogram Based method scores the firm (SAHOL-31.12.2013) with lower the current and cash ratio, net profit/net
sales, net profit/total asset, short-term liabilities/total assets than the mean values as the most prominent outlier
52
Page
53
Page
Peer-reviewed Academic Journal published by SSBFNET with respect to copyright holders.
Emir, et al. / International Journal of Finance & Banking Studies
Vol 4, No 4, 2015 ISSN: 2147-4486
Outlier
Date Firm V1 V2 V3 V4 V5 V6 V7 V8 V9
Score
31.12.2013 SAHOL 0,70 0,15 8.457.484.000 1.731.396.000 206.570.819.000 150.872.625.000 0,21 0,01 0,73 9,25
30.12.2011 KCHOL 0,80 0,13 68.969.387.000 2.124.469.000 98.621.087.000 62.031.471.000 0,03 0,02 0,63 8,96
30.12.2011 SAHOL 0,60 0,08 10.595.650.000 1.877.987.000 150.855.400.000 113.432.029.000 0,18 0,01 0,75 8,65
31.12.2014 SAHOL 0,65 0,09 10.517.510.000 2.079.114.000 230.526.555.000 169.096.825.000 0,20 0,01 0,73 8,55
31.12.2012 SAHOL 0,67 0,10 7.088.379.000 1.858.491.000 171.826.030.000 125.832.866.000 0,26 0,01 0,73 8,51
31.12.2012 KCHOL 1,37 0,45 65.449.383.000 2.324.150.000 48.687.778.000 15.816.621.000 0,04 0,05 0,33 8,40
31.12.2014 KCHOL 1,33 0,52 68.336.365.000 2.710.145.000 63.941.008.000 19.016.744.000 0,04 0,04 0,30 8,22
31.12.2013 KCHOL 1,35 0,47 65.942.213.000 2.679.713.000 58.789.960.000 19.182.420.000 0,04 0,05 0,33 8,10
31.12.2014 VKGYO 237,26 232,81 2.603.345 21.854.652 749.969.806 1.344.933 8,40 0,03 0,00 7,46
31.12.2013 TUPRS 0,94 0,35 41.078.427.000 1.197.223.000 21.139.387.000 10.396.466.000 0,03 0,06 0,49 7,19
31.12.2014 THYAO 0,77 0,20 24.157.801.405 1.819.259.536 31.875.607.062 8.505.344.748 0,08 0,06 0,27 6,29
31.12.2013 ALGYO 120,77 100,79 32.208.298 84.385.039 388.760.710 1.671.142 2,62 0,22 0,00 6,18
31.12.2014 KOZAL 17,41 14,34 885.888.000 494.890.000 2.019.552.000 74.667.000 0,56 0,25 0,04 5,77
31.12.2013 THYAO 0,68 0,21 18.776.784.325 682.707.427 25.402.077.818 6.652.755.831 0,04 0,03 0,26 5,71
30.12.2011 VKGYO 89,45 88,78 29.464.403 25.790.717 189.197.017 543.191 0,88 0,14 0,00 5,51
31.12.2012 TUPRS 1,13 0,40 42.436.908.000 1.464.119.000 16.647.907.000 8.068.100.000 0,04 0,09 0,49 5,49
31.12.2014 TUPRS 0,82 0,46 39.722.712.000 1.458.963.000 21.932.560.000 8.561.001.000 0,04 0,07 0,39 5,33
30.12.2011 TUPRS 1,08 0,15 40.747.047.000 1.241.738.000 14.757.986.000 8.386.151.000 0,03 0,08 0,57 5,25
31.12.2014 ECZYT 38,60 36,39 1.268.808.655 25.458.248 1.276.624.569 6.277.694 0,02 0,02 0,01 5,22
31.12.2014 ALGYO 75,00 63,56 14.919.333 78.151.728 463.942.210 2.994.025 5,24 0,17 0,01 5,21
54
Page
Peer-reviewed Academic Journal published by SSBFNET with respect to copyright holders.
Emir, et al. / International Journal of Finance & Banking Studies
Vol 4, No 4, 2015 ISSN: 2147-4486
Outlier
Date Firm V1 V2 V3 V4 V5 V6 V7 V8 V9
Score
31.12.2014 VKGYO 237,26 232,81 2.603.345 21.854.652 749.969.806 1.344.933 8,40 0,03 0,00 42,20
31.12.2013 ALGYO 120,77 100,79 32.208.298 84.385.039 388.760.710 1.671.142 2,62 0,22 0,00 18,58
31.12.2014 SAHOL 0,65 0,09 10.517.510.000 2.079.114.000 230.526.555.000 169.096.825.000 0,20 0,01 0,73 17,86
31.12.2013 SAHOL 0,70 0,15 8.457.484.000 1.731.396.000 206.570.819.000 150.872.625.000 0,21 0,01 0,73 14,39
31.12.2014 GOZDE 3,27 1,19 51.315 60.171.733 1.596.993.791 159.960.907 1172,60 0,04 0,10 10,70
31.12.2012 SAHOL 0,67 0,10 7.088.379.000 1.858.491.000 171.826.030.000 125.832.866.000 0,26 0,01 0,73 9,67
31.12.2013 BJKAS 0,08 0,01 142.106.681 -143.902.222 104.646.652 317.904.163 -1,01 -1,38 3,04 9,47
30.12.2011 SAHOL 0,60 0,08 10.595.650.000 1.877.987.000 150.855.400.000 113.432.029.000 0,18 0,01 0,75 7,85
30.12.2011 BJKAS 0,08 0,01 153.475.240 -150.801.385 126.947.414 345.134.942 -0,98 -1,19 2,72 7,30
30.12.2011 KCHOL 0,80 0,13 68.969.387.000 2.124.469.000 98.621.087.000 62.031.471.000 0,03 0,02 0,63 6,65
30.12.2011 VKGYO 89,45 88,78 29.464.403 25.790.717 189.197.017 543.191 0,88 0,14 0,00 6,21
31.12.2014 ALGYO 75,00 63,56 14.919.333 78.151.728 463.942.210 2.994.025 5,24 0,17 0,01 5,67
31.12.2014 KCHOL 1,33 0,52 68.336.365.000 2.710.145.000 63.941.008.000 19.016.744.000 0,04 0,04 0,30 5,66
31.12.2013 KCHOL 1,35 0,47 65.942.213.000 2.679.713.000 58.789.960.000 19.182.420.000 0,04 0,05 0,33 4,42
31.12.2012 KCHOL 1,37 0,45 65.449.383.000 2.324.150.000 48.687.778.000 15.816.621.000 0,04 0,05 0,33 4,08
31.12.2013 VKGYO 70,87 70,51 2.338.085 4.358.389 198.991.299 785.622 1,86 0,02 0,00 3,99
31.12.2012 BJKAS 0,15 0,05 147.426.786 -68.299.282 100.698.378 211.934.015 -0,46 -0,68 2,11 3,97
31.12.2014 AEFES 1,78 0,62 10.079.137.000 -512.233.000 20.113.805.000 2.533.723.000 -0,05 -0,03 0,13 3,16
31.12.2014 BJKAS 0,21 0,00 222.818.633 -140.478.649 179.621.502 338.397.664 -0,63 -0,78 1,88 3,01
30.12.2011 TUPRS 1,08 0,15 40.747.047.000 1.241.738.000 14.757.986.000 8.386.151.000 0,03 0,08 0,57 2,86
55
Page
Peer-reviewed Academic Journal published by SSBFNET with respect to copyright holders.
Emir, et al. / International Journal of Finance & Banking Studies
Vol 4, No 4, 2015 ISSN: 2147-4486
The top 20 outlier results of Robust PCA Outlier Detection algorithm are given in Table 7. The top outlier observation
is (VKGYO - 31.12-2014) while it is the ninth in histogram-based outlier list. Outliers do not have consecutive orders
as histogram-based approach. In the list, there are eight different firms. The first outlier is the firm having the highest
current and cash ratios among all observations. The second outlier (ALGYO-31.12.2013) has values of current and cash
ratios far greater than the mean. GOZDE and BJKAS join as outliers firstly in this list. (GOZDE-31.12.2014) is at
forefront of net profit/net sales while (BJKAS-31.12.2013) has the highest value of short-term liabilities / total assets.
Table 8 demonstrates the firm based summary of outlier methods. There are 15 different firms in the three lists produced
by three outlier detection methods. KCHOL and SAHOL have the great number of appearances with 12 observations.
KCHOL, SAHOL and TUPRS are firms that appears on all lists. EKGYO, BIMAS, TCELL, YAZIC for k-NN global;
ECZYT, KOZAL for histogram-based; BJKAS, GOZDE for robust PCA are placed in its own list only.
Empirical findings show that each outlier detection approach uses variables in different weights. Thus, methods give
different outlier firm lists to explore the inner structure of data set.
Conclusion
In this study, it is aimed to identify abnormal observations with respect to the fundamental analysis. Within this scope,
three outlier detection algorithms namely k-NN Global Outlier Score , Histogram - Based Outlier Score and Robust
PCA Outlier Score are utilized using end year values of nine significant fundamental analysis variables of firms listed
in Borsa Istanbul for the period between 2011 and 2014.
According to the results of k-NN Global Outlier Score algorithm, nine different firms took place in the outlier list.
(SAHOL-31.12.2014) has the maximum values of the variables total assets and short term liabilities. Additionally,
KCHOL and TUPRS have the next four places consecutively. In addition, EKGYO, YAZIC, BIMAS, TCELL and
AEFES are the other outliers in the list.
(SAHOL-31.12.2013) is also the first outlier of Histogram - Based Outlier Score algorithm. It has lower current and
cash ratios, net profit/net sales, net profit/total asset, short-term liabilities/total assets comparing with others. However,
Histogram Based method does not produce a consecutive outlier order structure as in k-NN Global algorithm. KCHOL,
56
(VKGYO - 31.12-2014) is the top outlier in Robust PCA Outlier Score algorithm whereas it is the ninth outlier in
Histogram-Based outlier list. Similar to this situation, other outliers do not also have consecutive orders as histogram-
57
Page
References
Agrawal, S., & Agrawal, J. (2015). Survey on Anomaly Detection using Data Mining Techniques. Procedia Computer
Science, 60, 708713. http://doi.org/10.1016/j.procs.2015.08.220.
Aizenman, J., Pinto, B., & Radziwill, A. (2004). Sources for Financing Domestic Capital-Is Foreign Saving a Viable
Option for Developing Countries? (No. w10624). National Bureau of Economic Research.
Amer, M., & Goldstein, M. (2012). Nearest neighbor and clustering based anomaly detection algorithms for rapidminer.
In Proc. of the 3rd RapidMiner Community Meeting and Conference (RCOMM 2012) (pp. 112).
Amsaveni, R., & Gomathi, S. (2013). Fundamental Analysis of Selected FMCG Companies in India. Asia-Pacific
Finance and Accounting Review, 1(3), 37.
Angiulli, F., & Pizzuti, C. (2002). Fast Outlier Detection in High Dimensional Spaces. In T. Elomaa, H. Mannila, & H.
Toivonen (Eds.), Principles of Data Mining and Knowledge Discovery (pp. 1527). Springer Berlin
Heidelberg.
Azhar Rosly, S., & Afandi Abu Bakar, M. (2003). Performance of Islamic and mainstream banks in
Malaysia. International Journal of Social Economics,30(12), 1249-1265.
Ben-Gal, I. (2005). Outlier Detection. In O. Maiomon & L. Rockach (Eds.), Data Mining and Knowledge Discovery
Handbook: A Complete Guide for Practitioners and Researchers. New York: Kluwer Academic Publishers.
Berthold, M. R., Borgelt, C., Hppner, F., & Klawonn, F. (2010). Guide to Intelligent Data Analysis: How to
Intelligently Make Sense of Real Data: Making Practical Sense of Real Data. Springer.
Bistrova, J., & Lace, N. (2015). Relevance of fundamental analysis on the Baltic equity market. Economics and
Management, (14), 132-137.
Brown, T. A., & Bukovinsky, D. M. (2001). ECR and grocery retailing: an exploratory financial statement
analysis. Journal of Business Logistics, 22(2), 77-90.
Callao, S., Jarne, J. I., & Lanez, J. A. (2007). Adoption of IFRS in Spain: Effect on the comparability and relevance of
financial reporting. Journal of International Accounting, Auditing and Taxation, 16(2), 148-178.
Carslaw, C. A., & Mills, J. R. (1991). Developing ratios for effective cash flow statement analysis. Journal of
Accountancy, 172(5), 63.
Chandola, V., Banerjee, A., & Kumar, V. (2009). Anomaly detection: A survey. ACM Computing Surveys, 41(3), 1
58. http://doi.org/10.1145/1541880.1541882.
Chapman, N. (2015). My Experience with Fundamental Analysis. Honors Scholar Theses. Paper 442.
Chu, T. C., Tsao, C. T., & Shiue, Y. R. (1996, December). Application of fuzzy multiple attribute decision making on
company analysis for stock selection. In Fuzzy Systems Symposium, 1996. Soft Computing in Intelligent
Systems and Information Processing, Proceedings of the 1996 Asian (pp. 509-514). IEEE.
Cooper, R., & Kaplan, R. S. (1991). Profit priorities from activity-based costing. Harvard Business Review, 69(3), 130-
135.
Costuleanu, C. L., & Codreanu, C. M. (2010). Statement of changes in equity. Publishing House of Iai University of
Agricultural Sciences and veterinary medicine, Scientific papers, series Agronomy, 53(2).
Damodaran, A. (1996). Corporate Finance. Wiley.
Edmister, R. O. (1972). An empirical test of financial ratio analysis for small business failure prediction. Journal of
Financial and Quantitative analysis, 7(02), 1477-1493.
Fama, E. F. (1998). Market efficiency, long-term returns, and behavioral finance. Journal of financial economics, 49(3),
58
283-306.
Page
Fan, H. F. (2015). An Empirical Study for Fundamental Analysis-Evidence from the Listed Electronic Corporations in
Taiwan.
Ranganatham, M. (2006). Investment analysis and portfolio management. Pearson Education India.
Page
RapidMiner Extension: Anomaly Detection (2014). German Research Center for Artificial Intelligence, DFKI GmbH.
Retrieved from http://madm.dfki.de/rapidminer/anomalydetection.
60
Page